1
SILICA SCALE MITIGATION FOR HIGH RECOVERY REVERSE OSMOSIS OF GROUNDWATER FOR A MINING PROCESS
P. Sanciolo a*, N. Milne a, K. Taylor b, M. Mullet b, S. Gray a
a: Victoria University, Melbourne, Australia
b: Hatch Associates, Perth, Australia
* Corresponding author
Abstract:
The feasibility of silica removal in RO treatment of groundwater from a Western Australian mining
and processing operation to prevent scaling and enhance water recovery was investigated. This
study has shown that it is possible to decrease the silica concentration in RO concentrate to levels
that would allow an overall water recovery of 90% to 95% using 10 g/L of regenerable activated
alumina adsorbent. Regeneration of the adsorbent using 2% NaOH was found to be effective for at
least three regeneration cycles. A preliminary costing of the high water recovery RO process using
silica removal by adsorption indicated product water (permeate) costs of $5.6/kL and savings due to
a reduction in brine volume from the current 40% of feed volume to 5‐10% of feed volume. It also
allows better utilization of a scarce groundwater resource, allowing the production of up to 1.6
times more low salt water from a given volume of groundwater. These results warrant larger scale
investigation of silica removal and adsorbent regeneration for high recovery RO processing for
mining operations, and application of silica removal to RO treatment of other silica laden waters
such as coal seam gas produced water.
Keywords: Silica, reverse osmosis, groundwater, scaling, high‐recovery RO.
2
1. Introduction
The quality and quantity of process water used in mining and mineral processing operations often
have a major impact on the efficiency of the process. In many remote mining operations in arid
regions, the only source of water is groundwater, and there is often a need to demineralize this by
reverse osmosis (RO) for process stages such as steam generation.
The presence of ionic components such as calcium, magnesium, carbonate, sulfate, phosphate and
silicate, and components such as soluble silica (silicic acid), in the feed water of an RO process can
limit the achievable water recovery by forming insoluble scale that hinders membrane filtration.
While it is possible to mitigate scale formation in RO processes by ionic components by removing
them using ion exchange resins [1] or accelerated seeded precipitation [2, 3], the mitigation of silica
scale remains a challenge which requires more research [4].
Silica scale mitigation in RO processes is a challenge because it is difficult to accurately predict scale
formation as the threshold limits of silica scale formation are influenced by a large number of
parameters [5]. Although it is possible to estimate silica scaling potential of a test water based on
the pure water solubility limits [6], the scaling potential of the water can be expected to be higher
than this prediction if even small quantities of trivalent metal ions are present [7] or if the ionic
strength is high [8], and lower than this prediction under conditions where the polymerization
reactions responsible for the formation of amorphous silica are slow, e.g., at low pH [8].
Silica scale mitigation is also a challenge because silica removal can lead to waste products that pose
engineering challenges for their management. It is possible, for example, to precipitate silica in the
presence of metal ions such as Fe(III) and Al(III) [9], but the resulting precipitate is gel‐like and
difficult to dewater [10]. It is also possible to adsorb the silica onto metal oxides such as alumina,
goethite and hematite [11, 12, 13]. If these are colloidal, however, this too can lead to sludge
management problems. In this paper, silica removal by adsorption onto activated alumina is studied.
Activated alumina was chosen due to its effectiveness as an adsorbent for silica [11,14] and the
tendency of the adsorbed silica to form a soft, gel‐like, layer around the adsorbent [15 ], thus
potentially facilitating adsorbent regeneration. In contrast, silica adsorption has been found to form
hard, glass like, layers around iron oxide adsorbents [16]. The removal of silica by various
concentrations of activated alumina, at different pH values and at different temperatures was
studied. The feasibility of regeneration of the activated alumina with caustic rinsing was also
investigated.
3
2. Experimental
2.1 Materials
The test water used in these experiments was the RO concentrate that had been derived from the
60% RO water recovery of a Western Australian groundwater. Only one batch of test water was
used in all experiments, and the water was collected and transported without acidification or
preservation. The test water was sampled for complete analysis on receipt and stored at ambient
indoor conditions (~20°C) until needed (~ 2 weeks). All experiments were performed within a one
month period and no changes in water quality were detected during this period. The activated
alumina used in these experiments (Rio Tinto Alcan, nominal specific surface area: 250‐270 m2/g,
d50<15 µm) was used as supplied.
2.2 Chemical Analyses
Initial chemical characterization of the RO concentrate (ROC) test brine was performed by a
commercial analytical laboratory. Subsequent silica, calcium and magnesium analyses were
performed by Inductively Coupled Plasma spectroscopy. The pH measurements were performed
using a Hach pH meter with gel filled double junction reference electrode and built‐in temperature
sensor.
2.3 Scaling Potential
The RO scaling potential was estimated for 90% water recovery of the ROC using ROSA8 Dow Filmtec
Software. The maximum silica concentration above which the silica precipitates in the ROC was
determined experimentally using a flat sheet Sterlitech Sepa CF RO module, operated with full
recycle of the concentrate such that the water recovery increased with time during the experiment,
similar to the procedure of Jawor 2009 [17] and Zach‐Maor 2008 [18]. The concentrate reservoir and
its content were maintained at 20°C by immersion in a constant temperature chilled water bath. The
membranes were wet before being loaded into the flat sheet module by immersion in deionised
water overnight. Samples of the concentrate were taken, filtered, and analysed for silica, calcium
and magnesium by ICP. The operating conditions for the flat sheet RO experiments are shown in
Table 1. The clean water flux of the membranes was measured over a 20 minute period immediately
prior to the RO concentrate scaling experiment. The average clean water flux for the different
membranes was found to be 25.3 ± 7.1 L/m2/h.
4
Table 1: Operating conditions used in scaling study (0.635 mm shim, 1.2 mm feed spacer
Parameter Value/type
Membrane type SW30‐ULE
Membrane area (m2) 0.014
Concentrate flow (L/min) 2.0
Temperature (°C) 20
Approximate initial permeate flow (L/min) 0.005
Approximate initial flux (L/m2/h) 21
Cross flow velocity (m/s) 0.28
Applied pressure (kPa) 2,500
2.4 Silica removal
2.4.1 Optimization of adsorption conditions
All experiments were performed using 100 mL of RO concentrate in stirred beakers. An IKA RCT Basic
(IKAMAG) magnetic stirrer with integrated temperature control was used. Experiments investigated
the effect of pH with activated alumina at a dose of 2 g/L, and with 15 minute contact time. The
effect of contact time was investigated at 20°C and 45°C and at 2 g/L and 10 g/L, at native pH (8.6).
The effect of temperature was investigated at native pH, 2 g/L adsorbent dose, at 20°C, 35°C and
45°C. The test solution was sampled and filtered through 0.45 micron cellulose acetate syringe filter,
and acidified with 3 drops of concentrated nitric acid prior to ICP analysis.
2.4.2 Activated Alumina Regeneration
2.4.2.1 Loading of silica onto adsorbent
The activated alumina was recycled through several adsorption steps before regeneration to better
utilise the capacity of the activated alumina. Such processing would also be used industrially to
minimise the activated alumina and adsorbent regeneration requirements for the process.
Unused alumina was loaded with silica from the ROC by recycling each batch of alumina through a
series of adsorption steps, so that the adsorption capacity of the alumina could be approached.
Loading was performed in a series of batch 100 mL adsorption steps at the set adsorbent
concentration of 10 g/L, at 25 ⁰C, and with 60 minute contact time. In between each batch
adsorption step, the liquor was filtered through 0.45 micron cellulose acetate filter and analysed for
silica by ICP to determine the extent of loading. Loading was ceased when the residual silica
concentration was approximately 80% of the untreated ROC concentration, i.e. after 5 loading steps.
5
Loading of regenerated alumina were performed under the same conditions as the loading of
unused alumina, using 5 silica loading steps.
The cumulative quantity of silica loaded onto the adsorbent (Lx) was calculated form the quantity of
silica on the adsorbent from the previous loading cycle (Lx‐1), the before and after adsorption
concentration (CB and CA), the volume of test solution (V), and the quantity of adsorbent used (M),
as shown in Equation 1.
. [1]
2.4.2.2 Adsorbent Regeneration
The loaded adsorbent was filtered through 0.45 micron cellulose acetate filter and then dispersed in
approx. 100 mL of 2% NaOH solution to give an adsorbent concentration of 10 g/L (assuming no
losses during loading stages), and stirred for 30 minutes. The alumina suspension resulting from this
step (Step 1) in the regeneration was filtered through a 0.45 micron cellulose acetate membrane
filter. A 15 mL sample of the filtrate was acidified with concentrated nitric acid and analysed for
silicon, aluminium, calcium and magnesium by ICP to determine how much silica has been removed,
how much of the alumina adsorbent had dissolved and if any calcium and magnesium had been
adsorbed and released during regeneration. The removed adsorbent was redispersed in 100 mL of
fresh 2% NaOH and the process repeated 8 times, giving a total of 9 regeneration steps. The alumina
remaining at the end of the 9th step was lightly rinsed with 5 mL of deionised water and the loading/
regeneration cycle was repeated two times. A diagrammatic representation of one
loading/regeneration cycle is shown in Figure 1.
The cumulative quantity of silica desorbed during regeneration (Dx) as a percentage of the silica
present on the adsorbent after step 5 in the loading cycle (L5, Equation 1) was calculated (as SiO2)
from the silicon concentration in the regenerant solution (CSi), the volume of regenerant used (V),
the weight of adsorbent (M), and the atomic/molecular weight of Si and SiO2 (28 and 60
respectively) as shown in Equation 2.
6
. .. 100 [2]
The cumulative quantity of adsorbent lost by dissolution during regeneration (ALx) as a percentage
of the added alumina (M) was calculated from the aluminium loss in the previous regeneration step
(ALx‐1), the concentration in the regenerant solution (CAl), the volume of regenerant (V), and the
atomic/molecular weight of Al and Al2O3 (27 and 102 respectively) as shown in Equation 3.
. .. 100 [3]
Figure 1: Summary of one loading‐regeneration cycle showing 5 loading steps (LS) and nine
regeneration steps (RS), using RO concentrate (ROC) and 2% NaOH regenerant.
100 mL NaOH
100 mL NaOH
100 mL NaOH
100 mL NaOH
100 mL NaOH
100 mL NaOH
100 mL NaOH
100 mL NaOH
100 mL NaOH
100 mL silica
depleted ROC
100 mL silica
depleted ROC
100 mL silica
depleted ROC
100 mL silica
depleted ROC
100 mL silica
depleted ROC
100 mL ROC
100 mL ROC
100 mL ROC
100 mL ROC
100 mL ROC
RS1 RS5RS2 RS3 RS4 RS6 RS7 RS8 RS9
LS1 LS2 LS3 LS4 LS51g
ActivatedAlumina
100 mL silica laden NaOH
100 mL silica laden NaOH
100 mL silica laden NaOH
100 mL silica laden NaOH
100 mL silica laden NaOH
100 mL silica laden NaOH
100 mL silica laden NaOH
100 mL silica laden NaOH
100 mL silica laden NaOH
7
3. Results and Discussion
3.1 Characterization of Minara Resources RO concentrate
The composition of the RO concentrate (ROC) generated at the mining site for an RO water recovery
of 60% is shown in Table 2.
Table 2: Chemical analysis of RO Concentrate, as performed by a commercial NATA, Australia, accredited laboratory.
Parameter Units Value
pH pH units 8.6
Reactive P (HL) - Phosphorus, reactive as P mg P / L 0.06
TKN/TP (HL) - Total Kjeldahl Nitrogen as N mg N / L 0.3
TOC - Total Organic Carbon mg/L 17
TDS at 180°C - Total Dissolved Solids mg/L 5800
Fluoride - Fluoride, as F mg/L 2.0
Chloride - Chloride, as Cl mg/L 2900
SO4 DA - Sulphate, as SO4 mg/L 810
Sulphide - Sulphide, total mg/L <0.1
Alkalinity - Bicarbonate Alkalinity as CaCO3 mg CaCO3 / L 460
Alkalinity - Carbonate Alkalinity as CaCO3 mg CaCO3 / L 54
Alkalinity - Hydroxide Alkalinity as CaCO3 mg CaCO3 / L <2
Alkalinity - Total Alkalinity as CaCO3 mg CaCO3 / L 510
TCN - Total Nitrogen as N (Calc) mg/L 42
NH3 as N (DA) - Ammonia, as N mg N / L <0.1
React. Silica - Silica, reactive as SiO2 mg/L 160
MS Total Metals - Aluminium mg/L 0.7
MS Total Metals - Barium mg/L 0.10
MS Total Metals - Boron mg/L 3.0
MS Total Metals - Cobalt mg/L <0.01
MS Total Metals - Copper mg/L <0.01
MS Total Metals - Iron mg/L <0.2
MS Total Metals - Lead mg/L <0.01
MS Total Metals - Manganese mg/L <0.01
MS Total Metals - Nickel mg/L <0.01
MS Total Metals - Strontium mg/L 3.0
MS Total Metals - Zinc mg/L <0.01
OES Scan - Calcium mg/L 280
OES Scan - Magnesium mg/L 230
OES Scan - Potassium mg/L 73
OES Scan - Sodium mg/L 1300
OES Scan - Acid soluble Si, as SiO2 mg/L 160
8
It can be seen from Table 2 that the main cationic constituents of the brine were Na, Ca and Mg. The
main anionic constituents were Cl, SO4 and CO3. The silica content of the brine was 160 mg/L and
this was totally in the reactive (non‐polymeric) form as both the reactive silica and acid soluble silica
concentrations were the same.
3.2 . Scaling potential of ROC
The scaling potential of the test brine (Feed) and a 90% water recovery RO concentrate with respect
to calcium carbonate, calcium sulfate and silica, as predicted by the Dow Filmtec Rosa 8.0 modelling
software, is summarised in Table 3.
Table 3: ROSA8 Dow Filmtec Software Scaling Information with respect to the major potential scale
constituents of the test water at the native pH of 8.6 and 90% water recovery.
LSI CaSO4 Saturation (%) Silica Saturation (%)
Feed Concentrate Feed Concentrate Feed Concentrate
1.8 4.8 16 250 81 445
It can be seen in Table 3 that the test brine (ROC) had a high potential for CaCO3 scale formation, as
indicated by the high positive values of the Langelier Saturation Index (LSI). The potential for CaSO4
scale formation was also high. The use of lower pH and antiscalants is recommended for RO
treatment of this water. The CaSO4 saturation is slightly above the conservative upper limit of 230 ‐
400% of saturation recommended by membrane manufacturers [19], indicating that calcium
removal may also be needed if higher water recoveries than 90% are required. The software also
predicts that a 90% water recovery will give rise to silica scale formation. This assessment, however,
is based on the silica solubility in pure water and does not take into account kinetic factors that can
give rise to different solubility behaviours in complex solutions [8, 20].
The silica, calcium and magnesium concentrations in the samples taken during the flat sheet RO
scaling tests are shown in Figure 2. The silica concentration (Figure 2a) shows that the silica
concentration remained at approximately 170 mg/L, indicating that the silica had precipitated to its
solubility limit. Comparison of this figure with the starting silica concentration of 160 mg/L reveals
that the initial RO concentrate was close to this silica solubility limit before the start of the
experiment. The calcium concentrations (Figure 2b) do not significantly deviate from the calculated
no‐precipitation line, suggesting that calcium did not precipitate. This is in contrast to the ROSA
modelling which is based on equilibrium solubility data alone, which showed an LSI of 4.8 for the
9
concentrate, which is conducive for precipitation [21]. This may be attributable to the large
quantities of magnesium present, which can act as a powerful inhibitor of calcium carbonate
precipitation [22]. The magnesium concentration (Figure 2c) also does not deviate from the no‐
precipitation line at water recoveries less than 80%. For magnesium, however, there seems to be a
downward deviation at water recoveries greater than approximately 82%, suggesting that the
magnesium may be involved with the silica precipitation at water recoveries higher than
approximately 82%. The ROSA modelling did not indicate that the magnesium concentration was
conducive to magnesium hydroxide scale formation (results not shown).
Figure 2: Silica, calcium and magnesium concentration as a function of water recovery, solid line
represents the expected concentration in the absence of scale formation, different symbols are the
results obtained on replicate experiments.
The approximately 170 mg/L silica solubility limit indicated in Figure 2a allows estimation of the
target initial silica concentrations required to achieve different water recoveries in the RO treatment
of ROC, as shown in Figure 3. It can be seen from Figure 3 that if, for example, a 40 mg/L residual
silica concentration could be achieved, this would allow a 78% water recovery in the second stage
treatment of ROC before reaching 170 mg/L silica, equating to an overall water recovery of 92%, i.e.,
60% (the water recovery achieved in the production of the ROC test water) + 78% of the remaining
40%.
0
200
400
600
800
1000
1200
1400
1600
50 60 70 80 90 100
Silica Concentration (mg/L)
(a)
50 60 70 80 90 100
Calcium Concentration (mg/L)
Water recovery (%)
(b)
50 60 70 80 90 100
Magnesium Concentration (mg/L)
(c)
10
Figure 3: Prediction of target silica concentrations to achieve different water recoveries in the
second RO stage (after silica removal). The lines represent the expected silica concentration in the
absence of scale formation.
Although these scaling studies illustrate the difficulty involved in prediction of scaling conditions, it is
noteworthy that the conditions of the batch mode flat sheet experiments with concentrate
recirculation used in the current study are different to those prevailing in conventional industrial
scale spiral wound continuous RO process where the concentrate enters and leaves the membrane
system in a much shorter time. Slow precipitation of brine constituents may not lead to scaling in the
continuous process because the brine may have left the membrane system by the time the
precipitation takes place. In the recirculation mode experiments, however, where typically 10 to 12
hours were required for the desired water recovery to be achieved, scale formation from slowly
precipitating constituents is more likely. The batch mode recirculation mode experiments are,
therefore, potentially more vulnerable to scale formation from slow precipitation than continuous
mode experiments and provide a more conservative estimation of scaling propensity. Clearly, these
batch‐made recirculation mode studies are not a substitute for on‐site scaling studies at the
proposed design conditions.
0
20
40
60
80
100
120
140
160
180
200
50 60 70 80 90 100
Silica Concentration (mg/L)
Water recovery (%)
Initial Conc = 10 mg/L
Initial Conc = 20 mg/L
Initial Conc = 30 mg/L
Initial Conc = 40 mg/L
Initial Conc = 50 mg/L
Initial Conc = 60 mg/L
170 mg/L silica
solubility limit for this
test water
11
3.3 Silica removal by adsorption
3.3.1 Introduction
In order to achieve a system water recovery of 90% or above, a minimum of 75% water recovery
must be achievable on the remaining concentrate from the primary RO process. As can be seen in
Figure 3, the target silica concentration for the feed water to the second RO stage required to
achieve these water recoveries for the ROC is approximately 45 mg/L.
3.3.2 Effect of pH
The precipitation behaviour of silica is not as well understood as that of other scale forming
chemicals such as calcium carbonate or calcium sulfate [14]. In the absence of multivalent metal ions
it is believed to involve the polymerization of silicic acid (Si(OH)4) to form amorphous silica [23]. In
the presence of multivalent metal ions, however, the formation of metal silicates is possible at pH
values close to the pKa of silicic acid (9.8) [23]. The ROC used in these experiments contained
sufficient quantities of calcium, magnesium and aluminium for silicate scale formation to occur at
alkaline pH values [7]. The presence of calcium and magnesium have also been found to accelerate
polymerization of silicic acid at acidic pH values, but the mechanism for this is not clearly understood
[9]. This effect was found to be particularly pronounced at low silica concentrations (~100 mg/L).
The effect of pH on the concentration of soluble silica, calcium and magnesium, after contact with
activated alumina for 15 minutes at ambient temperature is compared to the silica, calcium and
magnesium concentration in the absence of adsorbent in Figure 4. It can be seen that in the absence
of adsorbents, the silica concentration decreases at pH values higher than the native pH of the ROC
(8.6). The calcium and magnesium concentration data indicates that the large decrease in silica
concentration that occurs between pH 8.6 and pH 9.6 in the absence of adsorbent does not involve
large decreases in calcium or magnesium concentrations, suggesting the silica concentration
decreases is due to polymerization of the silicic acid to form insoluble polymeric silica in this pH
range. As the pH is increased past the first pKa of silicic acid (9.8 at 25°C), the silica, calcium and
magnesium concentrations all decrease considerably, suggesting the formation of calcium and/or
magnesium silicates due to the increasing concentration of silicate ion, and/or the formation of
calcium carbonate and magnesium hydroxide at these higher pH values.
Comparison of the data in the presence and in the absence of activated alumina reveals that the
maximum decreases in silica concentration over what is achieved in the absence of adsorbent
occurred at the native solution pH of 8.6. The calcium concentration decrease of 34 mg/L (0.85 mM)
12
observed at this pH suggests the formation of calcium silicate. The silica concentration decrease of
44 mg/L (0.46 mM) suggests that at least some of the calcium precipitation does not involve silica.
The observed adsorption maximum is in agreement with the results of Bouguerra et al (2007) [11]
which showed a maximum adsorption of silica onto activated alumina at approximately pH 8. These
authors performed a regression analysis of the equilibrium data and found that the data fitted both
Langmuir and Freundlich adsorption isotherms, suggesting that the adsorption process involves
monolayer sorption onto a surface with a finite number of identical sites. They hypothesised that
silica adsorbs as the silicate anion which increases in concentration as the pH is increased towards
and past the pKa of silicic acid, and the decrease in adsorption at pH values higher than
approximately 8 was due to electrostatic repulsion between adsorbent and adsorbate resulting from
the generation of negative surface sites on the activated alumina at these pH values. Similarly, Bond
et al (2007) [14] argued that the decrease in adsorption at high pH was due to competition for anion
adsorption sites by hydroxyl ions.
3.3.3 Effect of adsorbent dose, temperatures and contact time
The effect of activated alumina adsorbent dose at various temperatures is shown in Figure 5a. It can
be seen that at 2 g/L, an increase in temperature from 20°C to 45°C has very little effect on silica
removal. These results are consistent with the results of Matson et al (1981) [24] who found that an
increase in temperature up to 50°C did not significantly increase silica adsorption onto activated
alumina at a dose of 2 g/L. It can, however, be seen that the use of a higher temperature (45°C) with
a higher alumina doses (10 g/L) with a long contact time (60 minutes), is advantageous, allowing the
achievement of silica concentration well below the target value of approximately 45 mg/L or the use
of shorter contact time. The use of elevated temperatures is an attractive option for mining
operations where the process produces large quantities of waste heat that can be used to elevate
the temperature of the RO concentrate. This would, however, require the location of the silica
removal process near the source of waste heat (e.g. near the sulphuric acid leach autoclaves in the
mining process).
The effect of extension of the contact time to one hour at various temperatures and adsorbent
doses is shown in Figure 5b. It can be seen that extension of the contact time to one hour makes
little difference at low adsorbent dose, indicating that the alumina adsorbent surface is saturated
with silica at low adsorbent dose. Surface saturation at low adsorbent dose is confirmed by the high
adsorbent dose data which shows considerable further removal of silica. It can also be seen that the
target silica concentration for 90% water recovery (45 mg/L) can be achieved at high dose and low
temperature with a one hour contact time, but that higher temperatures and adsorbent doses are
13
required if shorter contact times are to be used. If higher water recoveries are required, the use of
high adsorbent dose (10 g/L) and high temperatures (45°C) and contact time are needed, giving a
silica concentration of approximately 15 mg/L and a water recovery of 96% (90% in the second RO
stage, see Figure 3).
Figure 4: Residual silica, calcium and magnesium concentration after 15 min contact time with
various adsorbents at various pH vales, 2 g/L adsorbent dose, ambient temperature.
Figure5: Residual silica concentration at pH 8.6, Left: after 15 min contact time with 2 g/L AA101
adsorbent at various temperatures, Right: at contact times up to one hour, at various temperatures
and AA101 adsorbent doses.
4 5 6 7 8 9 10 11 12
Magnesium Concentration (mg/L)
pH
4 5 6 7 8 9 10 11 12Calcium Concentration (mg/L)
pH
AA101
No Adsorbent
0
50
100
150
200
250
300
350
400
4 5 6 7 8 9 10 11 12Silica Concentration (mg/L)
pH
(a) (c) (b)
0
50
100
150
200
250
0 5 10 15
Silica Concentration (mg/L)
Adsorbent Dose (g/L)
45 Deg.C 35 Deg.C 20 Deg. C
0
50
100
150
200
250
0 20 40 60 80Silica Concentration (mg/L)
Time (minutes)
2 g/L adsorbent, 20 deg. C
2 g/L adsorbent, 45 Deg.C
10 g/L adsorbent, 20 Deg.C
10 g/L adsorbent, 45 Deg.C
(b)(a)
14
3.3.4 Regeneration of activated alumina adsorbent
The kinetics of silica adsorption on activated alumina is such that high concentrations of activated
alumina (approx. 10g/L) were required to treat the RO concentrate water in a reasonable time (eg.
30‐60 minutes). The use of high concentrations of activated alumina, however, results in low
utilisation of the adsorption capacity of the activated alumina, so the activated alumina was recycled
through several adsorption steps to better utilise the capacity of the activated alumina. Such
processing would also be used industrially to improve the kinetics of adsorption and to minimise the
activated alumina and adsorbent regeneration requirements for the process.
The silica loaded during five consecutive additions and removal by filtration of 100 mL of RO
concentrate to 1.0 g of AA101 at ambient temperature (18°C) with 60 minute contact time, for the
virgin adsorbent (First Loading), once regenerated adsorbent (Second Loading) and twice
regenerated AA101 adsorbent (Third Loading) are shown in Figure 6a. This figure also includes a
second order polynomial fit trend line to allow estimation of the adsorption capacity of the alumina
before and after the regeneration cycles. The quantity of silica that came off the loaded adsorbent
during the regeneration steps is shown in Figure 6b. The quantity of aluminium lost from the
adsorbent during the regeneration is shown in Figure 6c. The release of silica, aluminium, calcium
and magnesium during regeneration is shown in Figures 7a to 7d respectively. It can be seen from
Figure 6a that the loading capacity of the alumina was approximately 50 mg SiO2/g alumina, and
that this increased to approximately 60 mg SiO2/g alumina after regeneration. Figure 5b shows that
approximately 90% removal of adsorbed silica was achieved in the first regeneration, and that this
was achieved after 4 regeneration steps. The second and third regenerations, however, were found
to require more regeneration steps. The quantity of alumina lost during each regeneration (Figure
6c) was found to be greatest during the first regeneration (~30%) and decreased dramatically in the
second and third regenerations (to ~2%).
These results are consistent with dissolution of small alumina particles in the first regeneration and
the formation of an adsorption layer during the second and third adsorption stages that is less
soluble than the silica loaded, non‐regenerated, alumina in the 2%NaOH regenerant. Such an
adsorption layer may, for example, be the result of reaction of silicate ions with aluminium ions
liberated from the alumina at alkaline conditions (Figure 7b) to form an aluminosilicate. This
aluminosilicate layer may initially be a better adsorbent than virgin alumina [25], and this may
explain the slightly better adsorbent performance during the second loading (Figure 6a), the lower
silica desorption (Figure 6b) and lower adsorbent loss during the second regeneration (Figure 6c).
This effect is, however, likely to eventually lead to poorer adsorption due to a decrease in the
specific surface area of the activated alumina substrate, possibly resulting from blocking of the pore
15
network by the adsorbed aluminosilicate. This may explain the slightly poorer adsorption
performance in the third loading than in the second loading (Figure 6a). Further research on a larger
scale would be required to optimise the regeneration process and determine the lifetime of the
adsorbent.
Figure 6: The quantity of silica loaded during loading cycles (a), quantity of silica desorbed
during regeneration cycles (b), quantity of alumina lost (as Al2O3) during regeneration (c)
An interesting feature of the analysis of the 2% NaOH solution during the various regeneration steps
and regeneration stages (Figure 7), is the presence of measurable quantities of calcium and
magnesium in the regenerant (Figures 7c and 7d respectively).
R² = 1.0000
R² = 0.9996
R² = 1.0000
0
20
40
60
80
0 2 4 6 8
Silica Load
ed (mg/g)
Loading Step
First Loading Second Loading Third Loading
0
20
40
60
80
100
0 2 4 6 8 10
Silica Desorbed (%)
Regeneration StepFirst Regeneration Second Regeneration Third Regeneration
0
10
20
30
40
0 2 4 6 8 10
Adsorbent Loss (%)
Regeneration Step
First Regeneration Second Regeneration Third Regeneration
(a)
(c)
(b)
16
Figure 7: Measured levels of silica (a), aluminium (b), calcium (c) and magnesium (d) in the
2% NaOH regenerant solution during regeneration.
The quantity of magnesium released during regeneration (Figure 7d) was larger in the first step and
decreased to nearly zero during subsequent steps for all three regeneration stages. This indicates
that magnesium was being adsorbed along with the silica during loading, and that it was being
0
50
100
150
200
250
0 2 4 6 8 10
Silica Concentration
(mg/L)
0
50
100
150
200
250
0 2 4 6 8 10
Aluminium
Concentration (mg/L)
0
2
4
6
8
10
0 2 4 6 8 10
Calcium Concentration
(mg/L)
0
2
4
6
8
10
0 2 4 6 8 10
Magnesium
Concentration (mg/L)
Regeneration Step
First Regeneration Second Regeneration Third Regeneration
(a)
(d)
(b)
(c)
17
adsorbed to the same extent and in a similar manner onto the virgin and regenerated alumina
surfaces. It also suggests that this adsorbed magnesium species is not evenly distributed throughout
the adsorbed silica layer, but is predominantly on the outer surface of the adsorbed layer since it
was easily removed in only one regeneration step. This behaviour is different to that observed for
calcium (Figure 7c). The quantity of calcium released during the first step of the three regeneration
stages was approximately the same for all three regeneration stages, and decreased markedly for
the second and subsequent steps during the first regeneration, as was the case for magnesium. This
pattern is, however, not seen in the second and third regeneration, where the calcium level remains
approximately the same for all the regeneration steps. This calcium desorption pattern suggests that
the adsorption of calcium is similar to that of magnesium during the first loading stage, but that
calcium is more evenly distributed throughout the silica adsorption layer than magnesium in the
second and third loading stages. This, in turn, suggests that at least some of the silica adsorbs as a
low solubility calcium silicate and/or calcium aluminium silicate during the second and third loading
stages. These findings and tentative explanations highlight the complexity of the chemical
interactions involved in the current system, the feedwater specificity of the outcome of adsorption
processes, and the need for a better understanding of chemistry involved in order to gain better
control over the process.
4. Preliminary Costing
The three RO concentrate treatment options considered for preliminary costing were those that
resulted in silica concentration below the target concentration for 90% system water recovery (45
mg/L, 75% water recovery in the SWRO second stage):
Option 1: The silica is removed by high pH precipitation in the absence of adsorbent, with
precipitate removal using a clarifier, the overflow being neutralised with acid before
processing through a continuous SWRO plant.
Option 2: The silica is adsorbed onto the alumina at native pH of the ROC, at ambient
temperature, with no pH adjustment before processing through a continuous SWRO plant.
The adsorbent is regenerated using 2% NaOH.
Option 3: As for Option 2, but at an elevated temperature (45°C), using waste heat from
processing plant
A diagrammatic representation of the process flows is shown in Figure 8. The designs are based on a
total permeate flow of 500 m3/hr. The assumptions used for the OPEX estimates are shown in Table
4. The estimated capital expenditure (CAPEX) and operating expenditure (OPEX) are summarized in
Table 5.
18
Figure 8: Block Flow Diagrams of costed treatment options, Option 1(a), Option2 (b), Option 3 (c).
High pH precipitation: This option has a significantly higher cost ($4M). It is included here, however,
for comparison with the adsorption options. The major equipment considered in this preliminary
costing for this option are a caustic dosing system including pH controller, a clarifier, a sulphuric acid
dosing system including pH controller to neutralise clarifier overflow prior to membrane plant, and
modifications to current piping systems. This option will have one effluent stream which is the
underflow from clarifier containing 22 kg/hr silica. Volume is estimated to be 200 l/hr. This would be
directed to the tailings storage facility. The chemical usage will mainly be caustic and sulphuric acid.
Caustic is used to raise pH in order to precipitate silica. Sulphuric acid is used after the silica has been
removed in order to neutralise the caustic prior to the SWRO plant. Estimated requirements are 50%
caustic solution at 300 l/hr and 98% concentrated sulphuric acid at 150 L/hr. The chemical usage is
estimated to be $25,000 per day which equates to $2/kL of total system permeate (500 m3/hr).
(a) (b) (c)
19
Table 4: Assumed parameters settings for preliminary operating cost estimates
Parameter Value Comments
System water recovery (%) 90 75% water recovery in second RO stage
Adsorption Capacity (mg SiO2/g AA) 50 See Figure 6(a)
Volume of concentrate to be treated (kL/hr) 222 Concentrate produced at 60% water recovery at 500 m3/hr permeate production
Required silica concentration decrease (mg/L) 115 Initial 160 mg/L to be reduced to 45 mg/L by adsorption
Adsorbent loss on 1st regeneration (%) 30 See Figure 6(c)
Adsorbent loss on subsequent regenerations (%) 2 See Figure 6(c)
Average adsorbent loss over 10 regeneration cycles (%) 6 Average over first 10 regeneration cycles
50% Caustic : AA requirement (L : kg) 18 Using 5 regeneration steps (see Figure 6)
Number of days between regenerations 5 Based on adsorption capacity and size of adsorption vessels
Number of times regenerant is reused 10 Estimated
Cost of adsorbent ($/kg) 1.2 Estimated
50% NaOH solution ($/L) 3 Estimated
98% H2SO4 ($/L) 1 Estimated
Activated alumina adsorption at ambient temperature: This option has a significantly lower capital
cost ($2.7M) than alkaline precipitation (Option 1). The concentrate from the existing membrane
plant is passed through a fixed bed of activated alumina. The silica is adsorbed onto the alumina
reducing the concentration to a level where an SWRO plant can achieve 75% recovery, and therefore
an overall 90% recovery. Once the activated alumina is fully loaded the silica can be washed off using
a 2% caustic solution. This will regenerate the alumina so it can be re‐used. Some alumina will also
be dissolved by the regeneration solution (average 6%, see Table 4) and will require replacement.
The major equipment required for this option are activated alumina adsorption vessels, surge tanks
for feed to and product water from adsorption vessels, automated regeneration system using in‐line
caustic dosing, and modifications to current piping systems. There is one effluent stream – the spent
regenerant stream containing redissolved silica and some alumina. The volume is estimated to be 23
m3/hr containing 2% caustic, 900 ppm silica and 620 ppm alumina. This stream would be discharged
to the current mine tailings storage facility. It is anticipated that there will be several adsorption
vessels in series and that it will be around 5 days between regenerations when water production is
at full capacity. Effluent production will occur during regeneration. Effluent will be less when water
production is lower. The major chemical use will be 22 m3/day of 50% caustic solution, and 740
20
kg/day of activated alumina at a total cost of $67,000 per day, equating to 5.6 $/kL of total system
permeate.
Activated alumina adsorption at 45°C: This option is very similar to Option 3, but the adsorption
takes place at 45 °C. A heat source is required to elevate the temperature of the feed stream prior to
the first adsorption vessel. Ideally, the low silica exit stream would then be used to provide some of
the heat to the feed stream via heat exchangers, so that a dedicated cooling medium is not required
in order to reduce the feed temperature to the RO plant. Subject to the final location of the
adsorption stage, the required heat may be available as waste heat from current steam generation
processes. For clarity, heating and cooling stages are not shown in Figure 8. The estimated capital
cost are lower than for Option 2 due to the smaller number of adsorption vessels ($2.2 M). The
chemical costs are expected to be the same as for Option 2 (5.6 $/kL of total system permeate).
Table 5: Preliminary Cost estimates of 90% water recovery RO operation involving silica removal
Option CAPEX ($M)
OPEX ($/kL of permeate)
1: High pH silica precipitation and removal using sedimentation.
4.0 2.0
2: Silica removal using activated alumina adsorbent at ambient temperature
2.7 5.6
3: Silica removal using activated alumina adsorbent at 45 °C
2.2 5.6
5. Conclusions
It was shown that it is possible to decrease the silica concentration in RO concentrate to levels that
would allow an overall water recovery of 90% or above using high concentrations (10 g/L) of
activated alumina adsorbent. Regeneration of the adsorbent using 2% NaOH as regenerant is
effective for at least three regeneration cycles. Analysis of the 2% NaOH regenerant solution after
use gave evidence that suggests that the alumina surface chemistry is modified by the regeneration
process. The regenerated alumina was found to be less soluble and have a higher adsorption
capacity than the un‐regenerated alumina. The liberation of adsorbed silica from the regenerated
alumina was also lower than for the un‐regenerated alumina. These changes in surface chemistry
were tentatively attributed to the formation of aluminium silicate on the alumina surface during
21
regeneration. Regeneration was also found to release considerable quantities of calcium, suggesting
the possibility of the formation of calcium silicate and/or calcium aluminium silicate on the
adsorbent surface. These findings highlight the complexity of the chemical interactions involved in
silica removal by adsorption, the feedwater specificity of the outcome of adsorption processes, and
the need for a better understanding of chemistry involved in order to gain better control over the
process.
Preliminary costing of the high recovery RO process with an added silica removal stage indicates that
this added stage leads to high product water costs ($5.6/kL), and is likely to significantly decreases
the brine management costs due to the reduction of the brine volume from the current 40% to 5%
of the feed volume. It also allows greater mineral production in situations where groundwater
availability is a limiting factor. These results warrant larger scale investigation of silica removal using
adsorption column, and adsorbent regeneration.
Acknowledgements
The authors acknowledge the financial support of the National Centre of Excellence in Desalination
Australia which is funded by the Australian Government through the Water for the Future initiative,
and the collaboration of Hatch Associates.
References
[1] J. Irving, and J.A. Dale., The development and application of Type III strong base anion exchange
resins (Purolite A‐555), Advances in Ion Exchange for Industry and Research. 1998, Royal Society
of Chemistry: London, UK. p. 221‐236.
[2] A.C. Rahardianto, B. McCool, and Y. Cohen, Accelerated desupersaturation of reverse osmosis
concentrate by chemically‐enhanced seeded precipitation, Desalination 264 (2010) 256–267.
[3] P. Sanciolo, L. Zou, S. Gray, G. Leslie, D. Stevens, Accelerated seeded precipitation pre‐treatment
of municipal wastewater to reduce scaling. Chemosphere, 72(2) (2008)243‐249.
[4] S. Salvador Cob, , C. Beaupin, , B. Hofs, M.M. Nederlof, D.J.H. Harmsen, E.R.Cornelissen, A.
Zwijnenburg, F.E. Genceli Guner, G.J. Witkamp, Silica and silicate precipitation as limiting factors
in high‐recovery reverse osmosis operations, Journal of Membrane Science 423–424 (2012) 1–10.
22
[5] R. Semiat, I. Sutzkover, D.Hasson, Technique for evaluating silica scaling and its inhibition in RO
desalting, Desalination 140 (2001) 181‐193
[6] Dow Tech Manual Excerpt:
http://msdssearch.dow.com/PublishedLiteratureDOWCOM/dh_003c/0901b8038003c01d.pdf?fil
epath=/609‐02022.pdf&fromPage=GetDoc
[7] C. J. Gabelich, W.R. Chen, T. I. Yun, B.M. Coffey, I.H. Suffet, The role of dissolved aluminum in
silica chemistry for membrane processes, Desalination 180 (2005) 307‐319.
[8] G.A. Icopini, S.L. Brantley, and P.J. Heaney, Kinetics of silica oligomerization and nanocolloid
formation as a function of pH and ionic strength at 25°C, Geochimica et Cosmochimica Acta,
69(2), (2005) 293–303,
[9] R. Sheikholeslami, I.S. Al‐ Mutaz, T. Koo, A. Young, Pretreatment and the effect of cations and
anions on prevention of silica fouling, Desalination 139 (2001) 83‐95.
[10] L.G. Dyer, W.R. Richmond, P.D. Fawell, Simulation of iron oxide/silica precipitation in the
paragoethite process for the removal of iron from acidic zinc leach solutions, Hydrometallurgy
119–120 (2012) 47–54.
[11] W. Bouguerra, M. Ben Sik Alib, B. Hamrounia, M. Dhahbib, Equilibrium and kinetic studies of
adsorption of silica onto activated alumina, Desalination 206 (2007) 141–146.
[12] T. Hiemstra, M. O. Barnett, , and W. H. van Riemsdijk, Interaction of silicic acid with goethite,
Journal of Colloid and Interface Science (2007) 310: 8‐17.
[13] J. A. McKeague, Silica in solution in the presence of soil and related material: Forms,
concentrations and reactions", PhD, 1962, Cornell University, New York City, USA.
[14] R. Bond, S. Veerapaneni, Zero liquid discharge for inland desalination, AWWA Research
Foundation Report, 2007.
[15] T. Yokoyama, T. Nakazato, and T. Tarutani, Polymerization of silicic acid adsorbed on iron(III)
hydroxide, Bulletin of the Chemical Society of Japan, 53 (1980) 850‐853.
[16] T. Yokoyama, and T. Tarutani, Adsorption of monosilicic acid on aluminium hydroxide and
formation of silica, Memoirs of the Faculty of Science, Kyushu University Series C. Chemistry, 13
(1981) 23‐26.
23
[17] A. Jawor, and E.M.V. Hoek, Effects of feed water temperature on inorganic fouling of brackish
water RO membranes, Desalination 235 (2009) 44–57.
[18] A. Zach‐Maor, R. Semiat, A. Rahardianto, Y. Cohen, S. Wilson, S.R. Gray, Diagnostic analysis of
RO desalting treated wastewater, Desalination 230 (2008) 239–247.
[19] J. Thompson, A. Rahardianto, H. Gu, M. Uchymiak, , A. Bartman, M. Hedrick, D. Lara, J. Cooper,
J. Faria, P. D. Christofides, Y. Cohen, Rapid field assessment of RO desalination of brackish
agricultural drainage water, Water Research 47 (2013) 2649‐2660.
[20] A. D. Bishop, and J. L. Bear, The thermodynamics and kinetics of the polymerization of silicic acid
in dilute aqueous solution, Thermochimica Acta, 3(5) (1972) 399‐409.
[21] A. Antony, J. H. Low, S. Gray, A. E. Childress, P. Le‐Clech, G. Leslie, Scale formation and control in
high pressure membrane water treatment systems: A review, Journal of Membrane Science, 383
(2011) 1– 16
[22] T. Chen, A. Neville, M. Yuan, Assessing the effect of Mg2+ on CaCO3 scale formation–bulk
precipitation and surface deposition, Journal of Crystal Growth 275 (2005) e1341–e1347.
[23] R. Sheikholeslami, J. Bright, Silica and metals removal by pretreatment to prevent fouling of
reverse osmosis membranes, Desalination 143 (2002) 255‐267.
[24] J.V. Matson, 1981. Industrial waste‐water reuse by selective silica removal over activated
alumina. U.S. Patent 4,276,180.
[25] T. Yokoyama, A. Ueda, K. Kato, K. Mogi, and S. Matsuo, A study of the alumina‐silica gel
adsorbent for the removal of silicic acid from geothermal water: Increase in adsorption capacity
of the adsorbent due to formation of amorphous aluminosilicate by adsorption of silicic acid,
Journal of Colloid and Interface Science 252 (2002) 1‐5.
Top Related